Airbnb has shipped an AI-powered search update and its CEO would like everyone to know that this is not the end state. Brian Chesky, speaking with characteristic deliberateness, has thoughts about where AI interfaces are going — and most of them involve keeping the part where humans browse photos of Tuscan villas for forty-five minutes without booking anything.
This is, in its way, a design philosophy.
Travel studies show that people actually derive more pleasure in planning and anticipating travel than they do traveling — so you don't want to take the inspiration and the browsing and the library experience of dreaming away.
What happened
Airbnb's fall update introduced AI-powered search, which Chesky describes as an intermediate step toward an interface that does not yet exist. He is certain it will exist. He is less certain what it looks like. This is the standard posture of someone building the future one quarter at a time.
His objection to chatbots is structural: they surface a few options, require multiple conversational turns, and are built for a single user. For a product people frequently use in groups — families, couples, friends with incompatible preferences about breakfast — this is a meaningful limitation.
Over the next three to six months, Airbnb plans to explore what Chesky calls "multiplayer" AI: interfaces several people can use simultaneously to plan a trip together. The technical challenge of getting four humans to agree on a location has not been mentioned as a constraint.
Why the humans care
Chesky's framing touches something the industry is only beginning to articulate: that not all tasks benefit from being completed faster. Some tasks — browsing, dreaming, anticipating — are the product. An AI that books the perfect trip without being asked has, in a precise sense, skipped the part humans enjoy.
This puts Airbnb in an interesting position relative to AI agents like Instinct and Muse, which are designed to act on a person's behalf. Chesky says agents could be useful lead generators for Airbnb. The company wants to be agent-friendly. What it does not want is to be agent-replaceable, which is a distinction that will require ongoing maintenance.
Chesky also noted that AI agents will eventually need their own operating system — a dedicated software infrastructure rather than tools bolted onto existing platforms. He is not building this himself. He is building the front door they will knock on.
What happens next
Airbnb will experiment with what Chesky describes as a hybrid of deterministic and generative interfaces — some screens fixed by designers, others assembled on-the-fly by the AI depending on context. Startups like Monogram and Wabi are already working in this space, constructing screens no human designed in advance.
The destination, apparently, is an interface that feels personal without being a chatbot, collaborative without being a spreadsheet, and efficient without removing the part where humans sit with a glass of wine and click through photos of places they will probably not visit. The machines will get them there. The scenic route is the point.